A large-language model-driven approach to chronic disease follow-up

Chronic disease follow-up is essential for adjusting treatment plans and improving patient outcomes, yet traditional methods often face challenges such as resource inefficiency, insufficient coverage, and poor personalization. Leveraging advancements in artificial intelligence, this study proposes a chronic disease follow-up approach driven by a large language model (LLM). The methodology includes fine-tuning a pre-trained LLM using chronic disease-specific datasets to enhance task-specific capabilities. Integration with Graph-based Retrieval Augmented Generation (GraphRAG)technology enables structured and non-structed data utilization, improving retrieval and response accuracy. The system design incorporates personalized patient interaction, semantic similarity methods, and real-time knowledge updates to enhance compliance and efficiency. However, limitations such as the lack of comprehensive structured data and domain-specific constraints are noted, with future work aimed at refining knowledge graphs and optimizing workflows for vertical field applications. This study provides a novel framework for intelligent chronic disease management in the era of large models.